Star Classification: A Deep Learning Approach for Identifying Binary and Exoplanet Stars

TL;DR

This paper introduces a deep learning method that uses feature extraction and wavelet transformation on star light curves to accurately classify stars as binary or exoplanet hosts, facilitating large-scale astronomical data analysis.

Contribution

The paper presents a novel deep learning approach with wavelet transformation for star classification, achieving high accuracy and efficient processing of large astronomical datasets.

Findings

Achieved 81.17% test accuracy in star classification.

Wavelet transformation improved training efficiency.

Method applicable to large datasets from space telescopes.

Abstract

We present a novel approach for classifying stars as binary or exoplanet using deep learning techniques. Our method utilizes feature extraction, wavelet transformation, and a neural network on the light curves of stars to achieve high-accuracy results. We have also compiled a dataset of binary and exoplanet stars for training and validation by cross-matching observations from multiple space-based telescopes with catalogs of known binary and exoplanet stars. The application of wavelet transformation on the light curves has reduced the number of data points and improved the training time. Our algorithm has shown exceptional performance, with a test accuracy of 81.17%. This method can be applied to large datasets from current and future space-based telescopes, providing an efficient and accurate way of classifying stars.